Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
2.0
-
None
-
None
-
Windows XP commons Math 2.0 jre 1.6.0.16
Description
I am trying to optimize this:
maximize: v
v <= a1*p1 + a2*p2 + a3*p3 + a4*p4
v <= b1*p1 + b2*p2 + b3*p3 + b4*p4
v <= c1*p1 + c2*p2 + c3*p3 + c4*p4
v <= d1*p1 + d2*p2 + d3*p3 + d4*p4
p1 + p2 + p3 + p4 = 1
where a1-d4 are constant specified below by the code (i didn't want to copy and paste them up here. you can look below to see what they are in the objective function).
LinearObjectiveFunction f = new LinearObjectiveFunction(
new double[]
, 0 );
Collection<LinearConstraint> constraints = new ArrayList<LinearConstraint> ();
constraints.add(new LinearConstraint(new double[]
{ -1, 1.7316145027890766, 1.3584341412980305, 0.9305633063383639, 1.687117394945513 },
Relationship.GEQ, 0));
constraints.add(new LinearConstraint(new double[]
{ -1, 0.6617060079461883, 1.4862459822191323, 0.7692647272328988, 0.7329140944025636 },
Relationship.GEQ, 0));
constraints.add(new LinearConstraint(new double[]
{ -1, 1.3255966888982322, 286.21607948837584, 1.135907611434458, 0.9803367440299271 },
Relationship.GEQ, 0));
constraints.add(new LinearConstraint(new double[]
{ -1, 0.5428682596573682, 1.5745685116536952, 1.4834419186882808, 1.2884923232048968 },
Relationship.GEQ, 0));
constraints.add(new LinearConstraint(new double[]
{0, 1, 1, 1, 1},
Relationship.EQ, 1));
RealPointValuePair solution = null;
try
catch (OptimizationException e)
{ e.printStackTrace(); }I get this error back from the SimplexSolver.
org.apache.commons.math.optimization.linear.NoFeasibleSolutionException: no feasible solution
at org.apache.commons.math.optimization.linear.SimplexSolver.solvePhase1(SimplexSolver.java:177)
at org.apache.commons.math.optimization.linear.SimplexSolver.doOptimize(SimplexSolver.java:187)
at org.apache.commons.math.optimization.linear.AbstractLinearOptimizer.optimize(AbstractLinearOptimizer.java:106)
at Runner.main(Runner.java:101)
One interesting thing to note is that if you round all the numbers to the nearest 100's place, it works. If you keep it with the floating point precision shown here, it doesn't.